Quantitative bronchoalveolar lavage (BAL) is used to diagnose ventilator-associated pneumonia (VAP). We prospectively compared semiquantitative (SQ) and quantitative (Qu) culture of BAL for VAP diagnosis. Ventilated patients suspected of VAP underwent bronchoscopic BAL. BAL fluid was examined by both Qu (colony-forming units [CFUs]/mL) and SQ culture (none, sparse, moderate, or heavy) and results were compared. VAP was defined as 10⁵ CFU/mL or greater on Qu culture. Over 36 months, 319 BALs were performed. Sixty-three of 319 (20%) showed diagnostic growth by Qu culture identifying a total of 81 organisms causing VAP. All 63 specimens showed growth of some organism(s) on SQ culture with 79 of 81 causative organisms identified and two (Pseudomonas, one; Corynebacterium, one) not identified. The remaining 256 specimens did not meet the threshold for VAP by the Qu method. Among these, 79 did not show any growth on SQ culture. Among the 240 specimens showing some growth on SQ culture, a total of 384 organisms were identified. VAP rates in relation to strength of growth on SQ culture were: sparse, 10 of 140 (7%); moderate, 24 of 147 (16%); and heavy, 45 of 97 (46%). Sensitivity (Sn), specificity (Sp), positive (PPV), and negative (NPV) predictive values of SQ culture of BAL fluid for the diagnosis of VAP were 97, 21, 21, and 97 per cent, respectively. Nonquantitative culture of BAL fluid is fairly accurate in ruling out VAP (high Sn and NPV). It however has poor Sp and PPV and using this method will lead to unnecessary antimicrobial use with its attendant complications of toxicity, cost, and resistance.